Despite its importance in many applications, anomaly detection using one-class classification techniques, is notorious for its inability to scale to large features and datasets. We offer __FROCC__ family of one-class classification techniques that are super-fast (12x fast) _and_ accurate (2 pp better than closest deep learning based method). Oh...it scales to millions of dimensions and training datasets easily!
Can we learn temporal models even with some missing observations? We offer an efficient neural point process named __IMTPP__ in this paper, which can accurately predict next events in a temporal event stream.
We use the time-series of medical events as the background information for linking diagnoses terms in a medical ontology. Accepted as a full paper at [ACM BCB 2019](http://acm-bcb.org).
Random walks strike again!! This time for solving regular simple path reachability. Our ARRIVAL paper accepted in SIGMOD 2019
Our paper extending the IJCAI’17 publication on location-specific influence modeling over LBSNs has been accepted in ACM Transactions on Intelligent Systems and Technology (ACM TIST), one of the top journals in data mining and information systems!
Our paper on using random-walks for index-free reachability on dynamic and temporal graphs accepted at ICDE 2019!